The bacterial protein Hfq: stable modifications and growth phase-dependent changes in SPAM profiles

Abstract

In bacteria transcription is coupled to translation, and while it is broadly accepted that transcription–translation complexes (TTCs) are formed in growing bacterial cells, the exact spatial organization of these macromolecular assemblies is not known with certainty. Recent studies indicated the formation of orderly cytosolic superstructures in growing E. coli cells. The bacterial nucleic acid (NA)-binding protein Hfq has been shown to function at the interface of RNA synthesis–degradation machinery; multiple, independent studies link Hfq to orderly cytosolic assemblies. In this work, using fast cell lysis/2D-PAGE and in vitro reconstitution analyses we studied the Hfq modifications and small protein-associated molecules (SPAM). We demonstrate that native Hfq carries stable modifications and simulate 2D patterns of native Hfq–SPAM complexes in reconstitution experiments with purified Hfq and synthetic NA probes. We also demonstrate that genetically engineered Hfq lacking the conserved arginine residues positioned near the rim of the disc formed by the subunits’ N-terminal domains binds DNA with a reduced affinity in comparison with wild-type Hfq. These results are consistent with the proposed Hfq-mediated DNA remodeling and point to the involvement of this patch of conserved arginines in interactions with DNA.

1. Introduction

In E. coli bacteria transcription is physically coupled to translation. A key contribution to our understanding of the phenomenon of transcription–translation coupling was arguably made by Oscar Miller and co-workers, who, over five decades ago, used electron microscopy (EM) to acquire a compelling visual evidence for the formation of massive DNA–RNA polymerase–mRNA–(ribosome)_n complexes [1]. However, to this day, the exact spatial organization of these macromolecular assemblies in growing E. coli cells remains uncertain, and from Miller’s research we know that the addition of detergents to cell extracts was key to obtaining orderly depositions of TTCs onto EM grids (which may point to a significance of hydrophobic interactions for the compaction of these macromolecular complexes in living cells). The notion that bacterial TTCs could be somehow compacted and/or scaffolded seems logical in view of the functional organization of the eukaryotic translation apparatus, which adopted the membranous surfaces of endoplasmic reticula (ER) as scaffolds for ribosomes. Fast-forward to more recent studies with E. coli (which introduced fluorescence microscopy [FM] for visualization of NA-binding proteins and nucleoid): it was first noted during the study of the E. coli nucleoid-associated proteins that growing cells harbor in their cytosolic region orderly supramolecular structures distinct from the core nucleoid. In the groundbreaking study by Azam et. al [2] the bacterial nucleoids – stained with 4′,6-diamidino-2-phenylindole (DAPI) – have been shown to co-exist with recognizable cytosolic foci, which were visualized through immunostaining with Hfq-specific antibodies [2] (a cartoon rendition/compilation of the images reported in [2] is shown in Fig. 5A below). More recently, the formation of morphologically similar structures was confirmed by other research groups [3,4]; the study by Taghbalout et al. mapped multiple proteins (including Hfq) to orderly cytosolic superstructures [3], which the authors refer to as the “organized cellular network of RNA processing and degradation machinery” [3]. Due to limitations of FM, it is not possible to tell with certainty whether the observed supramolecular structures represent large singular cisterns with protein or proteolipid surfaces – which could be comparable (functionally) to eukaryotic ER – or networks of smaller complexes [3,5,6], or both. Furthermore, because the DAPI staining (a staple technique for visualization of bacterial nucleoids) has DNA density-specific limitations, we cannot rule out that the structures in question could envelope the transcribed domains of the partially de-condensed nucleoid – as opposed to being stand-alone cytosolic assemblies. While alternative methods for visualization of the E. coli nucleoid DNA (based on the expression in bacterial cells of DNA-binding proteins fused to various fluorescent proteins) indicated the formation of ‘lobes’ in post-G1 nucleoids [8], the distinctly peripheral position of the Hfq-specific foci [2–4] and their partial overlap with the components of translation apparatus [2] would seem to differentiate these structures from nucleoid domains per se. Notwithstanding limited insights into the molecular organization of these complexes, a growing body of evidence indicates that during cell growth the bulk of Hfq is located in these orderly cytosolic structures (i.e. the cytosolic structures that possess non-random geometric organization).

It was proposed that the C-terminal domains of Hfq subunits or their modifications could play a role in interactions with cellular scaffolds [9,10], and there is some evidence that removal of this portion of the protein makes the cellular distribution of Hfq more uniform [4]. In accord, biochemical analyses of native RNA polymerase–(RNA)–accessory protein(s) complexes indicated that both RNA polymerase and Hfq are largely insoluble [9,11,12] and likely accessorized with lipids [9,11]; this could explain the formation of putative proteoliposomes under certain conditions in vitro (see Discussion) and, arguably, anchoring of these proteins in lipid or proteolipid monolayers in vivo. Localization of the bulk of cellular Hfq to poorly characterized cytosolic superstructures – rather than inner membrane or core nucleoid – is not broadly recognized and could have been overlooked in some studies. The question of the primary localization of Hfq, however, is key to understanding its explicit role(s), which have been shown to affect many processes in bacterial cells [13,14] and of significance to bacterial virulence [15]. Admittedly, the functional analyses of Hfq in refined model systems in vitro have proven to be challenging, in part, due to dissimilar NA-binding properties displayed by the purified native and recombinant Hfq proteins [9]; these differences may be due to the effect(s) of (artificial) protein modifications/tags [9] and/or protein-associated NA fragments that are likely present in Hfq obtained through certain purification methods. While the involvement of Hfq in RNA transactions and its perceived effects on the stability and maturation of biological molecules are often listed among Hfq functions (reviewed in [16–18]), it is not certain if these broadly manifested biological effects result from a remodeling of hypothetical NA substrates by soluble Hfq₆ complexes or mere removal or alteration of a largely insoluble structural component of the bacterial cell that may be involved in scaffolding or other forms of general maintenance of NAs; the former possibility seems increasingly unlikely due to an apparent lack or scarcity of soluble Hfq species [9,11,12].

Since the characterization of protein modifications and SPAM (small protein-associated molecules; the term reserved for non-covalently bound accessory molecules other than cofactors) could be instrumental to understanding the explicit catalytic role(s) of Hfq and likely could illuminate the mechanism(s) of its integration into orderly cytosolic superstructures, we sought to further study the Hfq modifications and SPAM through analytical methods dissimilar to those utilized in our prior work [9]. The results reported herein – which include the isoelectric focusing (IEF)-based separation of non-modified and modified Hfq forms, as well as native and in vitro-reconstituted Hfq–SPAM complexes – indicate, in accord with our prior findings, that Hfq carries post-translational modifications. We demonstrate growth phase-dependent changes in the makeup of Hfq SPAM and the formation of double-copy (dodecameric) Hfq assemblies in actively growing E. coli cells. We also tested new techniques for reconstitution of certain low-affinity Hfq–NA complexes. Using side-by-side comparative analyses of wild-type Hfq and genetically engineered Hfq with mutations along the projected DNA binding path, we studied the reconstituted protein–NA complexes in a refined system in vitro. Our data are consistent with Hfq-mediated DNA remodeling and point to the Hfq subunits’ peripherally positioned clusters of conserved arginines as sites that could contribute to the binding of DNA.

2. Materials and Methods

2.1. Bacterial strains

E. coli cell extracts were prepared using the E. coli MG1655 strain. Cells were grown at 37°C in LB Broth (Sigma-Aldrich L3022) using New Brunswick Scientific C76 water bath shaker; the harvesting conditions are specified in the text and Figure Legends. Recombinant (full-length, non-tagged) Hfq was expressed in E. coli M15/pRep4 cells (Qiagen) transformed with the plasmid pQE60Hfq^WT [12] or pQE60Hfq^R16P,R17G (see below). Typically, 1–2 L of E. coli cell culture (in LB Broth supplemented with 100 μg/ml Ampicillin and 25 μg/ml Kanamycin) was used per individual protein purification procedure; cells were harvested 2 hours after induction with 1 mM IPTG.

2.2. Preparation of cell extracts and protein purification

‘Fast’ cell lysis protocol was derived from the previously described procedure [12]; the protocol is described schematically in Fig. 1A. The extracts were either immediately utilized for 2D-PAGE experiments or, alternatively, flash-frozen on dry ice and stored at −75°C; a partial fragmentation of native dodecameric Hfq(–NA) assemblies was typically observed in cell extracts subjected to repeated cycles of freezing–thawing, as previously reported [12].

Figure 1. Growth phase-dependent changes in the levels of Hfq, RNA polymerase, and selected RNA polymerase-associated proteins.

A. Schematic of the cell lysis protocol that was used in the experiments described below. B. Determination of the levels of Hfq, RNA polymerase, and selected reference proteins in whole-cell extracts derived from E. coli cells harvested during active growth (AU₆₀₀=0.3; lanes 1 and 4), stationary phase (lanes 2 and 5), and late stationary phase (lanes 3 and 6; stationary and late stationary phase cells were harvested, respectively, 24 h and 48 h after the AU₆₀₀ reference point of 0.3). C. Quantitated results of the experiment described in Fig. 1B. Solid-color and dotted columns refer to the samples obtained through cell lysis protocols carried out, respectively, in the absence or presence of RNase A. Data show a representative result of two independent sets of experiments. Note that the RNase treatment moderately enhanced the extraction of certain RNA-binding proteins.

A representative purification procedure of recombinant (full-length, non-tagged) Hfq is described in Fig. S1; the protocol is loosely based on the previously described purification method [9]. Cell lysis in the presence of lysozyme and deoxycholate (DOC)-based protein extraction were carried out as previously described [9], except that the incubation of the cell suspensions with lysozyme (Sigma-Aldrich L6876) was shortened to 20 min, and the polyethylenimine (PEI, also referred to as Polymin P) fractionation was omitted. The chromatography stages of the modified protocol were also similar to those described in [9], except that DNA–Sepharose 4B chromatography was substituted with Source Q chromatography (Fig. S1); this modification was implemented to minimize a potential loss of modified Hfq species [9] during the initial purification step.

2.3. Two-dimensional polyacrylamide gel electrophoresis (2D-PAGE) and immunoblotting (IB)

Isoelectric focusing (IEF) was carried out as previously described [11], using BioRad PROTEAN^® IEF Cell and 7-cm, pH 3–10 nonlinear Immobilized pH Gradient (IPG) strips (BioRad). Unless indicated otherwise in Figure Legends, heat treatments of cell extracts and reconstituted protein–NA mixtures were omitted. The second (SDS-PAGE) stage was carried out using Mini-PROTEAN^® II Cell (BioRad) largely according to the manufacturer’s manual, except that custom-made Teflon combs with a side mini-well for Precision Plus Dual Color Standards (BioRad) were used. Following the IEF stage, the 2D-fractionated native proteins and protein–NA complexes were transferred onto Immobilon^®-P membranes (Millipore), typically, for 48–60 hours at 80 V (in a cold room), in 1x Tris Glycine Transfer Buffer, pH 8.3 (TGTB) supplemented with 10% HPLC-grade methanol. TGTB was obtained as a 10x stock from KD Medical. Immunostaining of the membranes with Hfq-specific antibodies [9] and chemiluminescent detection and quantitation of the protein were carried out as previously described [9]. Each 2D-PAGE/IB experiment was repeated at least twice, with varied modes of sample pre-treatment (as seen in Figs. 2A and 2B).

Figure 2. Identification of the Hfq modifications and SPAM by 2D-PAGE.

Preparation of E. coli whole cell extracts through ‘fast’ cell lysis protocols and 2D-PAGE/Immunoblotting (IB) were carried out as described in Materials and Methods. Monomeric and oligomeric Hfq complexes are highlighted with orange arrowheads. A. Identification of the Hfq modifications and SPAM in E. coli extracts. Hfq^M1-M3: modified Hfq species or Hfq–SPAM complexes. The Hfq–SPAM (putative Hfq–NA) complexes were loosely divided into two groups (M2, M3) based of the results of reconstitution experiments described in Fig. 3. Note that the Hfq–SPAM complexes were sensitive to heat treatments in the presence of excess SDS (as seen in Fig. 2B); however, under semi-denaturing conditions, these stable complexes could be fractionated by IEF. B. Identification of stable Hfq modifications in E. coli extracts denatured in the presence of excess SDS. This panel shows a representative 2D gel resulting from an experiment similar to that described in Fig. 2A, except that the E. coli cell extract was subjected to a 20-min, 80°C heat treatment in the presence of 2% SDS. The stability of the Hfq^M1 species indicates a post-translational modification of the protein and is consistent with the results of LC-MS analyses of purified native Hfq [9]. C. Quantitated levels of the Hfq and Hfq^M1 species observed in (SDS-denatured) 2D-PAGE-fractionated whole cell extracts of E. coli cells harvested during active growth. Data represent the results of four independent sets of experiments. This extent of modification suggests that, on average, 2–3 subunits per hexameric Hfq complex carry stable modifications.

2.4. Site-directed mutagenesis

Site-directed mutagenesis of E. coli Hfq was carried out using Q5^® Site-Directed Mutagenesis Kit (New England Biolabs). To engineer the Hfq R16P, R17G mutation, the ‘back-to-back’ mutagenic primers HFQRRPGM1 (5’-GTGAACGCACTGCCCGGGGAACGTGTT), HFQRRPGM2 (5’-GAACGGATCTTGTAAAGATTGCCC), and the template DNA pQE60-Hfq^WT [12] were used. The wild-type and mutant Hfq proteins (respectively, Hfq^WT and Hfq^R16P,R17G) were expressed in M15/pRep4 E. coli cells (Qiagen) and purified as described above. A complete DNA sequence of the insert region of the pQE60-Hfq^R16P,R17G construct is shown in Fig. S2.

2.5. Electrophoretic Mobility Shift Assays (EMSA)

EMSA experiments were carried out in a manner similar to that described in [9], typically, using 8% ProtoGel^™–1x TBE system. ProtoGel^™ was obtained from National Diagnostics; TBE was obtained as a 10x stock from KD Medical. The Alexa Fluor 488-labeled DNA probe was obtained from Invitrogen; the probe’s predicted secondary structure is shown in Fig. 4E. Following the electrophoresis in Mini-PROTEAN^® II Cell, gels were placed on High Performance Ultraviolet Transilluminator (UVP) and gel images were captured using Olympus C5060 camera outfitted with Vivitar Light Green Filter No. 11; little or no digital processing was applied to thus obtained gel images.

Figure 4. Effect of the removal of the conserved N-terminal arginine patch on Hfq–NA interactions.

A. A disc-shaped assembly formed by the N-terminal domains of subunits of wild type E. coli Hfq [47]. The conserved arginines located near the rim of the disc (R16, R17) are highlighted as red spheres. B. A homology model of the Hfq^{R16P, R17G} mutant. The structure was obtained using SWISS-MODEL homology modeling server [48]. C and D. Effect of the removal of the conserved arginine patch on electrophoretic mobility of Hfq and cross-linked Hfq–NA complexes. Purified full-length Hfq (obtained as described in Materials and Methods) was pre-incubated for 20 min at ambient temperature or, alternatively, at 80°C (the modes of sample pre-treatment are indicated in the figure) in the presence or absence of indicated NAs in Reaction Buffer D. When indicated, homopolymeric RNA (Poly[rA], GE Healthcare) was present at a final concentration of 5 A₂₆₀ units. Following the pre-incubation (and, when indicated, the Ribonuclease A treatment), the proteins and/or protein–NA mixtures were subjected to SDS-PAGE/Western Blot. Note the formation of cross-linked protein–NA complexes in heat-treated Hfq–NA mixtures; these complexes can be instantly recognized due to their altered electrophoretic mobilities. See text for details. E. Effect of the removal of the conserved rim arginines on Hfq–DNA interactions. EMSA utilizing the Alexa Fluor 488-labeled DNA probe (whose secondary structure is shown in the left panel) was carried out as described in Materials and Methods. With wild type Hfq, half-maximal binding of the indicated DNA probe was observed at 0.8–1 μM Hfq; this is in-line with the previously reported DNA-binding affinity of full-length Hfq [9]. Note that the knockout mutation of the N-terminal arginine patch resulted in Hfq’s reduced DNA-binding affinity (Fig. 4E, graph; also, Fig. 4E, gel, compare lanes 2–4 with lanes 6–8). A representative result of three independent experiments is shown.

3. Results

3.1. Growth phase-dependent changes in the levels of RNA polymerase (RP) and its main accessory proteins

First, we sought to determine the overall levels of Hfq, RP, and selected reference proteins (which included several RNA polymerase-associated proteins identified in our prior studies [19–21]) under the growth conditions utilized in the fast cell lysis/2D-PAGE experiments described below. To this aim, the previously developed fast cell lysis protocols ([11,12]; also, Fig. 1A) were modified to include cell extract obtained from E. coli cells harvested during active growth (AU₆₀₀=0.3) and stationary phase (Figs. 1B and 1C). In addition, in this set of experiments we tested the effect of RNase on the efficiency of extraction of several RNA-binding proteins accessory to E. coli RP, including Hfq. As seen in Fig. 1B, the transition from active growth to late stationary phase resulted in a near-complete shutdown of synthesis and/or apparent disposal of RP and some of its key accessory proteins, such as RpoD (Figs. 1B and 1C). In contrast, Hfq was present in stationary phase E. coli cells; the levels of Hfq were only moderately reduced during the transition from active growth to stationary phase (Fig. 1C). Interestingly, the addition of RNase A to lysis buffers resulted in moderately (7–12%) higher levels of the extracted S1 and RP (Fig. 1C). The solubility of a number of reference proteins remained largely unaffected in the presence of RNase A (Figs. 1B and 1C). The double-copy (dodecameric) Hfq₆–Hfq₆ complexes were consistently detected in actively growing E. coli cells (Fig. 1B, lanes 1 and 4), in accord with the previously reported data [9,11,12].

3.2. Hfq modifications and growth phase-dependent changes in SPAM profiles

In our prior studies, we developed fast cell lysis protocols with the purpose of assessing the composition of SPAM in minimally adulterated native E. coli RP–(RNA)–accessory protein(s) complexes [11,12]. While our prior work suggested that Hfq may carry post-translational modifications [9], the challenges of in vitro analyses of Hfq–SPAM complexes [9,12] compelled us to adopt a more incremental approach to studies with Hfq, which warranted a confirmation of our prior findings through an independent analytical method. In this work we tested if the putative non-modified and modified Hfq forms [9] could be partitioned by IEF.

In E. coli whole-cell extracts obtained through fast cell lysis protocols (Fig. 1A, schematic) and fractionated by 2D-PAGE, a broad palette of Hfq species carrying modifications and/or SPAM could be identified (Fig. 2A, Hfq^M1-M3 species). The putative Hfq species carrying stable modifications (Fig. 2A, Hfq^M1) could be differentiated from less stable Hfq–SPAM complexes (Fig. 2A, Hfq^M2-M3) based on sensitivity of the latter assemblies to heat treatments in the presence of excess SDS (compare Figs. 2A and 2B). In whole-cell extracts derived from E. coli cells harvested during active growth, in addition to monomeric Hfq species, Hfq complexes with electrophoretic mobilities consistent with those of hexameric and dodecameric Hfq complexes were also present (Fig. 2A, orange arrowheads). The dodecameric Hfq complexes were consistently detected in minimally diluted whole-cell extracts obtained from E. coli cells harvested during active growth; however, such assemblies were not seen in whole-cell extracts derived from early stationary-phase cells (Figs. 2A and S3; also, Fig. 1B, compare lanes 1 and 2, 4 and 5). This result was in accord with the previously reported data [11,12].

3.3. Reconstitution of full-length (non-tagged) Hfq–SPAM complexes in vitro

Since stable Hfq–SPAM (putative Hfq–NA) complexes were consistently seen in 2D-PAGE-fractionated whole-cell extracts derived from E. coli cells harvested during active growth (Hfq^M2 and Hfq^M3 forms in Fig. 2A), we wanted to demonstrate that complexes with similar IEF profiles could be reconstituted in vitro using purified Hfq and synthetic NA probes. To carry out these experiments, we purified recombinant (full-length, non-tagged) Hfq [12] using a variant of the previously described method (see Materials and Methods and Fig. S1). As seen in Fig. 3, the IEF profile of recombinant Hfq obtained through this method was comparable to that of native Hfq – minus the Hfq^M2, Hfq^M3 (putative Hfq–NA) forms.

Figure 3. 2D-PAGE patterns of native Hfq-SPAM complexes can be simulated using mixtures of purified Hfq and synthetic oligonucleotides.

The reference panel shows the 2D-PAGE profile of purified recombinant Hfq (obtained as described in Materials and Methods); a part of the panel shown in Fig. 2A is also included as a reference. To carry out the reconstitution experiments, purified recombinant Hfq was incubated with the indicated synthetic RNA oligonucleotides (obtained from Thermo Scientific Dharmacon^®) at 37°C for 20 min in 1x Reaction Buffer D (20 mM Tris-acetate, 10 mM magnesium acetate, 50 mM potassium acetate, 1 mM dithiothreitol, pH 7.9). Following the incubation, protein–NA mixtures were subjected to 2D-PAGE; 2D-PAGE and subsequent immunostaining with Hfq-specific antibodies were carried out as described in Materials and Methods. Interpretative schematics below identify individual protein–RNA complexes; see text for details.

Since the ability of Hfq to form high-affinity complexes with RNAs is well documented [22–24], we hypothesized that the Hfq^M2 and Hfq^M3 species could be such (detergent- and/or chaotropic agent-resistant) protein–RNA assemblies. To confirm their identity, we reconstituted Hfq–RNA complexes in vitro using purified (full-length, non-tagged) Hfq and synthetic RNA oligonucleotides of varied length and subjected the resulting protein–RNA mixtures to 2D-PAGE (with subsequent elution of the 2D-PAGE-fractionated material onto Immobilon-P membranes and immunostaining with Hfq-specific antibodies). As seen in Fig. 3, in these reconstitution experiments we were able to mimic the IEF patterns of native Hfq–NA complexes seen with 2D-PAGE-fractionated whole cell E. coli extracts. Specifically, the IEF patterns of the Hfq^M2 forms could be mimicked best using Hfq–(rA)₅ mixtures, while the IEF profiles of the Hfq^M3 forms (likely, stable Hfq[–NA] assemblies carrying longer NA fragments) could be simulated using Hfq–(rA)₂₀ or Hfq–(U)₂₀ mixtures (Fig. 3).

3.4. Effect of the removal of the conserved N-terminal arginine patch on Hfq–NA interactions

Prior studies with full-length E. coli Hfq indicated that it has a relatively low (μM- or near μM-range) affinity to dsDNA [25, 9]. (Higher DNA-binding affinities were reported for C-terminally truncated Hfq [26]). Archeal homologs of Hfq were also reported to bind supercoiled DNA substrates [27]. Recent research indicated that chemically synthesized polypeptides with sequences matching those found in the Hfq C-termini can self-assemble into amyloid fibers and that these assemblies can engage DNA [28–31]. Based on the results of Intact Protein MS analyses of purified native Hfq complexes, we proposed that alternating C-terminal polypeptides originating from adjacent Hfq hexamers could form a compound NA-binding scaffold [9,12]. In this study, we sought to develop techniques allowing to stabilize certain (low-affinity) Hfq–NA complexes, which could be instrumental to identification of the amino acids involved in interactions with NAs. Since it is generally known that NA fragments can be covalently cross-linked or otherwise tightly bound to multi-subunit protein complexes (for example, in our recent work we describe the identification of such fragments in protein–protein interfaces of the multi-subunit E. coli RP complex [11]), first, we tested the effect of heat treatments on Hfq–NA mixtures in a refined system, using polypurine NA probes and purified (full-length) Hfq. As an additional control for these protein–NA cross-linking experiments, we genetically engineered and purified Hfq^R16P,R17G mutant (Figs. 4B and S2; this mutation site was chosen because in our prior work we noted a loose, partial homology between Hfq and eukaryotic protein complexes that spool DNA, which suggested that this patch of conserved arginines could contribute to interactions with DNA [12]).

When (wild-type) Hfq–NA mixtures were subjected to heat treatments, invariably, SDS-resistant (Hfq₆)_n–NA assemblies (referred to below as cross-linked Hfq complexes) were formed (Figs. 4C, 4D, and data not shown). The role of NA in coupling of individual Hfq₆ complexes was evident from side-by-side comparative analyses, in which purified proteins were subjected to 80°C, 20-min treatments in the absence or in the presence of purified NA (in Fig. 4C, compare lanes 3 and 4 with lanes 5 and 6). The molecular masses of the cross-linked Hfq–NA species – which displayed electrophoretic mobilities dissimilar to that of non-modified Hfq – were consistent with those of complexes containing 2–4 Hfq hexamers per individual (protein–NA)_n assembly. Interestingly, while Hfq^R16P,R17G cross-linked to RNA with an efficiency comparable to that of wild-type Hfq (Fig. 4C, compare lanes 5 and 6), the nuclease trimming of the cross-linked protein–NA complexes formed by wild type and mutant Hfq indicated the formation of the adducts with dissimilar electrophoretic mobilities (Fig. 4D; this result points to a plausible role of the conserved arginine patch in protein–NA cross-linking). Co-incidentally, dissimilar electrophoretic mobilities of the Hfq^R16A mutant and wild-type Hfq complexes (obtained through a protocol that included a heat treatment of the crude cell extract) have been reported [32]. Next, we compared the DNA specificity of wild type and mutant Hfq side-by-side, using EMSA. These comparative analyses indicated that the removal of the conserved patch of arginines positioned near the rim of the disc-shaped assembly formed by the Hfq subunits’ N-terminal domains (Figs. 4A and 4B) resulted in Hfq’s reduced DNA-binding affinity (Fig. 4E).

4. Discussion

Multiple recent studies indicated that our understanding of the spatial organization of the bacterial transcription–translation apparatus may be flawed or incomplete [2–6]. While bacterial cells are often cartooned as single-compartment ‘sacks’ packed with DNA, RNA, and proteins, the shortcomings of the analytical methods which paved the way for these cartoons are rarely mentioned. These shortcomings include: (i) characteristically long processing times in imaging analyses of intact cells (in comparison with the remarkably fast dynamics of changes in post-G1 nucleoids [8]), (ii) reliance on prolonged cell lysis protocols and in vitro studies with diluted cell extract (in which low-affinity protein–protein and protein–NA assemblies undergo dissolution), and (iii) utilization of detergents and/or support grids with hydrophobic surfaces (which are certain to redistribute the types of molecules that play important roles in scaffolding of eukaryotic translation complexes). Taken together, these significant limitations of the previously applied analytical methods could have distorted the structure and composition of TTCs that are formed in living cells.

While the single-compartment models of bacterial cells that emerged as a result of utilization of ‘slow’ (or otherwise flawed) analytical techniques have been broadly disseminated, more recently, new methods have been developed that allowed to monitor the dynamics of relatively fast growth phase-specific changes in bacterial nucleoids [8,33]. Likewise, the utilization of imaging techniques more suitable for the detection of dynamic changes in bacterial TTCs led to the identification of orderly cytosolic superstructures in growing E. coli cells [2–6]. Since the latter superstructures were first described [2], only a modest progress has been made toward understanding of their molecular organization. (The limited resolution of FM and, indirectly, the transition from studies with native multi-protein complexes isolated from living cells to hypotheses-oriented research with individually expressed recombinant proteins – often truncated and/or lacking the post-translational modifications and/or functionally significant SPAM present in their native counterparts – likely contributed to the status quo). Due to our focus on studies with purified native E. coli transcription complexes [19,34–39] and their accessory proteins, which included Hfq [20,39], we sought to characterize the query proteins’ modifications and accessory molecules [9,11]. Since our prior research indicated that the RNA-binding proteins accessory to E. coli RP can form orderly superstructures (Fig. 5B and unpublished data) – which may be similar to the structures identified in super-resolution microscopic analyses of minimally adulterated intact cells [5,6] – we concluded that a systematic characterization of RP(–RNA)-associated proteins’ modifications and accessory molecules could be instrumental to better understanding of the molecular organization of such superstructures.

Figure 5. Localization of the E. coli protein Hfq to cytosolic foci and the formation of orderly superstructures by the RNA-binding proteins accessory to E. coli RNA polymerase.

A. Cartoon rendition/compilation of the results of immunofluorescence experiments reported in the groundbreaking work by Azam et al. [2]. The formation of superstructures morphologically similar to those reported by Azam et al. was independently confirmed [3,4]. H: recognizable Hfq-specific foci; N: core nucleoid; see text for details. B. Formation of orderly superstructures by the RNA-binding proteins accessory to E. coli RNA polymerase (M.V. Sukhodolets, unpublished data). Left panel: affinity chromatography-based purification of E. coli RNA polymerase and its accessory proteins (AP) through a variant of Purification Method A [11,19] allows for the separation of transcription complexes from accessory RNA-binding proteins [11,20,21,39]. This set of proteins includes Hfq [9,11,20], ribosomal proteins S1, S2, S6 [11,39] and some additional proteins (present in sub-stoichiometric amounts). Right panel: representative EM images of the orderly superstructures formed by the RNA-binding proteins accessory to E. coli RNA polymerase. The sensitivity of the reconstituted assemblies to high concentrations of glycerol and the recently reported solubility characteristics of some of the participant proteins [11,12] point to a significance of hydrophobic interactions for the formation of such complexes. See text for details. C. Hypothetical modes of scaffolding of E. coli transcription complexes. TC: transcription (DNA–RNA polymerase–nascent RNA) complexes; DNA and RNA are shown, respectively, in blue and green. For clarity, the RNA polymerase–RNA-associated proteins and ribosomes are omitted. SP_A: Scaffolding Protein(s) A; protein products of certain dna genes could arguably contribute to the formation of A-type assemblies. SP_B: Scaffolding Protein(s) B; the compound assemblies of this type may be synonymous to the orderly cytosolic superstructures identified in studies with intact bacterial cells [2–6]; see text for details.

4.1. Growth phase-dependent changes in the composition of oligomeric Hfq complexes

Our results confirm the formation of dodecameric Hfq complexes in actively growing E. coli cells [9,11,12]. Various modes of stacking of Hfq₆ complexes – resulting in formation of dodecameric assemblies – have been previously reported [40,41]. It seems plausible that the formation of such complexes could contribute to DNA remodeling in the nucleoid ‘awakened’ by the kickstart of RNA synthesis. Interestingly, unlike RP and some of the RP-associated proteins – whose levels dropped dramatically during late stationary phase – Hfq was consistently present in late stationary phase cells (in Fig. 1C, compare the growth phase-specific changes in levels of RpoB, RpoD, and Hfq), which may indicate its significance for ‘resting’ bacteria.

One of our objectives was to test the previously proposed models postulating a direct interaction between CTDs of adjacent Hfq hexamers during the formation of certain types of dodecameric Hfq complexes [9,12]. Since IEF allows for separation of non-modified and modified Hfq forms, our hypothesis – postulating that such complexes may be stabilized solely or predominantly by protein–protein interactions [9,12] – could be tested through comparisons of the focusing patters of the dodecameric Hfq complexes with those of non-modified (reference) Hfq and Hfq–SPAM complexes (including in vitro-reconstituted Hfq–NA complexes). The co-migration during IEF of the native hexameric and dodecameric Hfq complexes derived from minimally adulterated whole-cell extracts generated via fast cell lysis with the reference samples of purified, largely SPAM-free Hfq (Figs. 2 and 3) is consistent with our model.

4.2. Hfq modifications

Using IEF-based separation of the non-modified and modified Hfq forms, we demonstrate that Hfq – derived from either actively growing or stationary phase E. coli cells – carries post-translational modifications (Figs. 2, 3, and S3); the resistance of the Hfq^M1 species to heat treatments in the presence of excess SDS (Fig. 2B) suggests a stable modification of the protein. The presence of the protein species carrying additional negative charge(s) is in-line with the results of LC-MS analyses of purified native Hfq [9]. Furthermore, the two independent methods of analysis (LC-MS and IEF) indicated a similar extent of Hfq modification – 2–3 modified subunits per hexamer (Fig. 2C and [9]). Overall, IEF was a sensitive tool for measuring the extent of Hfq heterogeneity, and we recommend this method (as a low-cost alternative to LC-MS) for characterization of purified Hfq.

4.3. Growth phase-dependent changes in the composition of Hfq SPAM

We demonstrate growth phase-dependent changes in the makeup of Hfq SPAM. While actively growing cells harbored a plethora of Hfq–NA assemblies with broadly varying pI values – whose IEF patterns could be mimicked in reconstitution experiments using purified Hfq and synthetic RNAs of a varied length (Fig. 3) – little or no such complexes were detected in stationary phase cells (compare Figs. 2A and S3).

4.4. Reconstitution of (full-length, non-tagged) Hfq–NA complexes

We demonstrate that a knockout mutation of the conserved N-terminal arginine residues results in Hfq’s reduced affinity to DNA (Fig. 4E). The reduced RNA binding resulting from the Hfq rim arginine (R16A) mutation was previously reported (42). The efficiency of the heat treatment-induced Hfq^R16P,R17G–RNA cross-linking was comparable to that of wild-type Hfq (Figs. 4C and 4D). Given that the arginines in question are arranged in a circular pattern (near the rim of the disc-shaped assembly formed by the Hfq subunits’ N-terminal domains, Figs. 4A and 4B), this could imply interaction of the oligomeric protein complex with bent or looped DNA. Based on the aforementioned evidence – which is consistent with the reported compaction of DNA in the presence of Hfq [29] – and the previously reported data demonstrating that the native Hfq–NA–RP complexes (isolated from DNase-treated E. coli extracts) migrate as a single assembly during fast-flow gel filtration [9], we hypothesize that Hfq could function as a DNA loop-stabilizing component of TTC.

4.5. The function of Hfq

The ambiguities with Hfq’s cellular status and explicit role(s) in NA management are intricately linked to the gaps in our understanding of the spatial organization of TTCs in actively growing bacterial cells. This uncertainty likely has contributed to Hfq often being described as a ‘chaperone’ protein – the monicker applied broadly to enzymes with protein and NA substrates whose explicit catalytic roles in many, if not most cases could not be definitively determined. FM- and EM-based studies indicated a distinctly peripheral (but off inner membrane) localization of Hfq [2–4,43], likely, as an integral component of orderly cytosolic superstructures [2–4]. These data are in-line with the results of studies in which Hfq’s solubility and distribution in fractionated cell extracts were examined [9,11,12]. Based on (i) the partial overlap in the distribution of Hfq and certain components of translational apparatus [2] and (ii) the unambiguous evidence pointing to ‘soluble’ Hfq being a component of transcription complexes [9, Figs. 4 and 5 therein], it seems plausible that Hfq could contribute to the spatial alignment of TTCs. In principle, TTCs’ intermolecular entanglements and other types of nonproductive interactions could be reduced if they were tethered to protein or proteolipid surfaces/scaffolds (Fig. 5C). The tethering via (i) mRNA’s U-rich termini (with Hfq acting as a likely coupling factor [9]), (ii) selected, possibly, lipid-accessorized ribosomal proteins (the candidate proteins – which typically co-purify with RP–RNA complexes in fast purification protocols – are generally known [9,11,20]), (iii) lipid-anchored RNA polymerase and/or RNA [11] cannot be excluded. In actively growing bacterial cells, such scaffolding surfaces could envelope the components of transcription machinery – resulting in a full or partial compartmentation of the components of transcriptional apparatus (Fig. 5C, left panel). The existence of such flexible protein or proteolipid ‘sacks’ enveloping bacterial nucleoids is consistent with the evidence indicating that during active growth the chromosomal DNA does not expand beyond the central part/core of a cell [2,8], leaving the peripheral spaces to the translation apparatus [2]. Since protein–protein contacts are more ‘information-rich’ than lipid–lipid contacts, the likely strength and flexibility of such hypothetical protein or proteolipid envelopes could accommodate complex architectural changes seen in post-G1 E. coli nucleoids. When compared (functionally) to eukaryotic nuclear membranes, such flexible protein or proteolipid envelopes can be considered more ‘advanced’ means of cellular compartmentation of DNA. However, alternative models – with the nascent RNAs located outside protein or proteolipid cores (Fig. 5C, right panel) – also cannot be ruled out. Structurally, the latter assemblies could be vaguely reminiscent of the cores of certain RNA viruses, some of which are known to possess characteristic lattices with extruded, conformationally loose termini of the constituent proteins. These (B-type) superstructures could harbor the RNA phosphorolytic enzymes [3] and selected enzymes involved in metabolism of carbohydrates (that were previously thought to be located in ‘glycosomes’ [44]) and enable the recently proposed cytosolic liquid–liquid separation [45]. Based on the cumulative results of our in vitro studies with Hfq [9,12,20,46] and the results of studies with intact bacterial cells reported by other research groups [2–4], we propose that the explicit cellular role of Hfq is to enable the NA scaffolding and/or transport across orderly cytosolic superstructures. Such transport and/or remodeling of NAs during growth phase-specific transitions could allow for their proper (lower-entropy) spatial alignment during transcription–translation and RNA phosphorolysis; these processes may be attenuated by regulatory RNAs. The presence of Hfq in late stationary phase cells (Fig. 1) also points to its likely significance for the maintenance of chromosomal DNA in ‘resting’ cells. It seems plausible that both types of scaffolds (described schematically in Fig. 5C) could exist in bacterial cells, and some of the published studies – such as the referred to above work by Taghbalout et al. [3] – imply that bacterial cells may possess more than one type of a cytosolic superstructure and/or superstructures with varying protein makeups.

Bulletpoints.

• Multiple, independent studies link Hfq to orderly cytosolic superstructures

• Isoelectric focusing (IEF) was used to partition native non-modified and modified Hfq forms

• The IEF patterns of Hfq–SPAM complexes were simulated in in vitro reconstitution studies

• A knockout mutation of the Hfq rim arginines results in a reduced DNA specificity